How We Built a Flexible Vendor Management Platform Now Ready for White-Label Use

Story behind

The Unbench platform originated from a KITRUM internal initiative in 2019. Once, KITRUM’s CEO, Vlad Kytainyk, tasked the team with streamlining communication. He aimed to make specialist request exchange more efficient and comfortable for both internal teams and external partners. We initially built the tool on Google Sites as a simple prototype. The prototype quickly demonstrated its value, allowing KITRUM to publish requests and partners to submit candidates.

This internal success sparked external interest. Partners began requesting the same capabilities for their own talent needs. Organic demand prompted KITRUM to recognize Unbench’s broader market potential. The goal was to build a “memory layer” for companies. This layer would manage their project knowledge and vendor relationships effectively. Therefore, we rebuilt the platform from a basic internal tool into a comprehensive B2B solution that now has its team and processing.

Workflow

Unbench development followed an Agile-inspired approach. We emphasized rapid testing and validation. This methodology enabled us to adapt to the changing market and ensure our solution was as robust as possible.

Team Size & Members

The core development team consisted of a Product Owner, Business Analyst, QA, Team Lead/Senior Full Stack developer, Middle Full Stack developer, and a Product Designer. Leadership included:

• CEO — Julia Stalnaya
• Head of Platform — Alyona Honcharenko
• Head of Marketing — Olha Chura 

In addition, we invited external SEO experts for post-design optimization.

Task management tools

Communication Tools

Communication flowed through Slack, email, and regular Google Meet sessions (3–4 times per week). These meetings ensured continuous alignment and quick feedback loops.

Implementation Process

We prioritized building a flexible, scalable foundation. Product logic was frequently refactored to accommodate growing user needs and new business rules without compromising stability. For example, we developed the in-platform chat from scratch, as existing solutions lacked the necessary UX, flexibility, and seamless integration.

Deployment & Quality Assurance

We ensured rapid, reliable updates through a robust CI/CD pipeline via GitHub Actions. A staging environment mirrors production for thorough release testing. A feature flag system allows gradual rollouts, minimizing disruption.

Timeline

Initial core platform development spanned 3 months. The project remains ongoing, with continuous improvements and feature additions.

Solution: Unbench platform as a framework

Unbench evolved into a modular, scalable platform. The platform manages complex B2B relationships beyond simple hiring. Unbench functions as a robust framework, adaptable for various vendor/client management needs.

Discovery phase

The discovery phase rigorously identified the platform’s dual nature and core user flows (demand and supply). It analyzed key functional value propositions: verified companies, streamlined communication, and integrated internal chat. For the overall platform redesign, our research focused on efficient information organization and clean, minimalistic design principles, drawing inspiration from various CRM systems.

Unbench platform overview

Unbench works as an “operating layer.” The layer connects companies with verified partners across tech, marketing, and recruiting. Its core process — create request, moderation, validation, choose match, deal closing — is a highly efficient B2B matchmaking engine. This structured, automated workflow makes Unbench an ideal white-label framework for other B2B vendor/client management scenarios, such as:

• HR Solutions: Managing contingent workers, external recruiters, or specialized talent pools.
• Logistics: Connecting carriers with shippers, managing freight requests, and tracking service providers.
• Other Service Industries: Any B2B context requiring efficient matching, verification, and collaboration between service providers and clients. 

While Unbench is a flexible framework for white-label solutions, the platform does not currently offer public APIs for direct integration with external client-side systems. Instead, Unbench provides a robust core for building tailored solutions.

Scalable architecture

The team designed Unbench’s architecture to handle a rapidly growing user base, exemplified by its expansion to over 1,000 companies and 26,000+ users.

Unbench’s modular architecture makes adding new features or customizing existing modules straightforward for white-label clients. React components and backend routes are isolated and reusable. The GraphQL schema is extensible; custom mutations and queries are easy to add. We enable theming and UI customization via configuration and CSS variables, without altering core logic. This ensures rapid feature additions for new white-label instances with minimal risk to shared core logic.

MongoDB database uses a schema-flexible design, optimizing document relationships (users, companies, requests, candidates). Our developers indexed high-query fields (e.g., request status, company ID) and separated collections for candidate profiles, chat messages, and request metadata, referencing them efficiently. For large datasets, we used pagination and filtering to avoid loading unnecessary data. We also implemented efficient GraphQL resolvers with batching and lazy-loading. We leveraged horizontal scalability using AWS services. Asynchronous operations, such as candidate validation, run via background queues to maintain performance under varying loads.

Tech stack

Unbench uses a modern, flexible technology stack: React for the frontend, Node.js with Express.js for the backend, and GraphQL for efficient API querying. MongoDB serves as the database. 

The platform is hosted on Amazon Web Services (AWS). It uses EC2 for backend services, S3 for file storage (e.g., CVs, profile images), CloudFront for content delivery, and Elastic Load Balancing (ELB) for scalable traffic routing. MongoDB Atlas handles the managed database layer.